I am grateful for love, support and understanding from my dear husband Marius Fossmark and my loving parents Ann Elin Gilje and Dag Bergslien.

Transcription

1

2 ii

3 Abstract The Statfjord Field has entered a drainage strategy where the reservoir will be depleted such that gas liberates from the remaining oil in the reservoirs. Adequate modelling of vertical lift performance is needed to predict a realistic liquid offtake and thereby pressure-depletion rate. Wells in the Statfjord Formation have been producing from a gas cap in which some areas of the formation has disappeared. Production tests from wells located in such areas have been used as basis when analysing multiphase-flow correlations ability to model vertical lift performance. Calculations are done in Prosper, a well performance, design and optimization program developed by Petroleum Experts. Conceptual test data describing liquid and gas-condensate wells were sett up to study prediction of pressure drop, and differences between correlations. Measured downhole pressures from 203 production tests, from six wells located in the Statfjord Formation, were used to compare accuracy of correlations. Petroleum Experts, Petroleum Experts 2 and Petroleum Experts 3 were found to be the most accurate correlations, and were recommended to use when creating lift curves for the Statfjord full field model. A trend of too low pressures predicted at low gas-liquid ratio (GR), and too high pressures predicted at higher GR was observed. An attempt of making the correlation even more accurate for a wider gas-liquid range was done by tuning in Prosper. None of the attempted modifications gave increased accuracy for the whole GR range studied. It was proposed that modification to equations for liquid holdup, or in flow regime boundaries may improve accuracy over a wider GR range. A study of using tuned correlations and possible errors introduced when predicting future performance was performed. Only small errors were observed for a narrow GR range (as for the Brent Group, Sm 3 /Sm 3 ), and one correlation can be used for the entire time range. With higher gas-liquid ratios, errors introduced by using correlations tuned to test data may be significant, and it was recommended to change correlation as function of GR development. A recommendation of correlations to use and how they may be modified when predicting future performance of the Statfjord Field is included. iii

4 Acknowledgements There are several people I would like to express my gratitude towards in making of this thesis. First I gratefully acknowledge Statoil ASA and the University of Stavanger for allowing me to write this thesis. I would like to thank Kari Nordaas Kulkarni for posing the thesis and Svein Magne Skjæveland for his guidance and support as my faculty supervisor. I am heartily thankful to my supervisor, Øivind Fevang. He has encouraged, guided and supported me from the beginning to the final level, and through this enabled me to develop an understanding of the subject. I would like to show my gratitude to Håvard Thomassen auritsen for guidance, discussions and support. He has made the development of this thesis enjoyable and educational by being accessible and friendly. My good friends and fellow students, Dagny Håmsø and Kristian Grepstad, have been supportive and encouraging. We have helped and pushed each other through the process, especially when difficulties were encountered, and for this I owe them gratitude. I am grateful for love, support and understanding from my dear husband Marius Fossmark and my loving parents Ann Elin Gilje and Dag Bergslien. astly I would like to thank all of those who supported me in any respect during the completion of my thesis. Marthe Gilje Fossmark iv

5 Nomenclature A = pipe cross-sectional area C = liquid holdup parameter C = correction factor CN = corrected liquid viscosity number d = pipe inside diameter E k = dimensionless kinetic-energy pressure gradient f = correction factor f = friction factor f = no-slip volume fraction g = gravity H G = gas holdup H = liquid holdup = length B = bubble-slug boundary M = transition-mist boundary S = slug-transition boundary n = correction factor N D = dimensionless diameter number N Fr = Froude number N GV = dimensionless gas velocity number N = dimensionless liquid viscosity number N V = dimensionless liquid velocity number N Re = Reynolds number N v = dimensionless velocity number p = pressure q = volumetric flow rate R = superficial liquid/gas ratio S = slip ratio t = time v = velocity v b = bubble rise velocity v

6 Z = length Γ = liquid distribution coefficient Δ = difference ε = pseudo wall roughness factor ε/d = relative roughness ε = roughness variable θ = inclination angle from vertical λ = no-slip fraction µ = viscosity ρ = density σ = surface tension τ = shear stress ψ = correction factor Subscripts a = acceleration F = Fanning f = friction G = gas h = hydrostatic = liquid M = Moody m = mixture of liquid and gas n = no-slip o = oil S = superficial s = slip t = total TP = two-phase w = water vi

7 Abbreviations BB = Beggs and Brill DRm = Duns and Ros Modified DRo = Duns and Ros Original FB = Fancher and Brown Fm = Formation GR = Gas-liquid ratio Gm = Gray Modified GOR = Gas-oil ratio Gp = Group H3P = Hydro-3 Phase HB = Hagedorn and Brown MB = Mukherjee and Brill O = Orkiszewski O2P = OGAS 2.phase O3P = OGAS 3-phase O3PE = OGAS 3-phase Extended OD = Outer diameter P1 = Parameter 1, tuning parameter for hydrostatic gradient P2 = Parameter 2, tuning parameter fro the frictional gradient PE = Petroleum Experts PE2 = Petroleum Experts 2 PE3 = Petroleum Experts 3 PE4 = Petroleum Experts 4 PE5 = Petroleum Experts 5 PI = Productivity index THP = Tubing-head pressure THT = Tubing-head temperature VP = Vertical lift performance WCT = Water cut vii

8 Table of Contents 1 Introduction and Objectives Theory Single-Phase Flow Pressure-Gradient Equation Multiphase Flow Holdup Velocities Mixture-Fluid Properties Pressure-Gradient Equation Flow Regimes Calculation of Pressure-drop in ong Pipelines Pressure-Drop Correlations Fancher and Brown Correlation (Fancher and Brown 1963) Gray Correlation (Gray 1974) Hagedorn and Brown Correlation (Hagedorn and Brown 1965) Duns and Ros Correlation (Duns and Ros 1963) Orkiszewski Correlation (Orkiszewski 1967) Beggs and Brill Correlation (Beggs and Brill 1973) Petroleum Experts Correlations (Petroleum Experts 2010) Study of Pressure-Drop Prediction in iquid and Gas-Condensate Wells iquid Wells Gas-Condensate Wells Effects of Increasing Gas-Rate on Pressure-Drop Prediction Effects of Varying Oil-Water Ratio on Pressure-Drop Prediction Conclusions Comparison of Measured and Predicted Bottomhole Pressures Accuracy of Correlations Effect of Input Data on Accuracy of Correlations Conclusions Modification of Correlations to Match Measured Bottomhole Pressures VP Matching Method with Prosper (Petroleum Experts 2010) Effect of Tuning Correlations to Test Data Studying Manual Tuning in Prosper Conclusions Effect of Using Tuned Correlations in Simulations Simulation with ProdPot Simulations with Tuned Correlations ow GR Development High GR Development Sensitivity Analysis Conclusions Main Conclusions and Recommendations Recommendations Sources of Error References...95 Appendix A Description of Wells...98 Appendix B Figures viii

9 1 Introduction and Objectives Calculation of pressure drop in oil and gas wells will be important for cost effective design and of well completions and production optimization (Persad 2005). Flow up the tubing will usually be multiphase. Gas and liquid tend to separate and will normally not travel with the same velocities. Both temperature and pressure conditions will change in upwards multiphase flow. Calculation of pressure drop will thereby not be straight forward (Time 2009). Nevertheless, accurate prediction of pressure drop in oil and gas wells is needed to forecast well deliverability and to optimize depletion (Reinicke et al. 1987). Many multiphase flow correlations are proposed. Still, none of them are proven to give good results for all conditions that may occur when producing hydrocarbons (Pucknell et al. 1993). Analyze of available correlations are often the best way to determine which one to use (Brill and Mukherjee 1999). Some will be good for liquid wells, whereas others for gas. Most of the correlations available are to some degree empirical and will thereby be limited to conditions of which the correlations are based on (Pucknell et al. 1993). The Statfjord Field has entered a drainage strategy where the reservoir will be depleted such that gas liberates from the remaining oil in the reservoirs. There are two reservoirs, the Statfjord Formation and the Brent Group, exposed to this strategy. Most of the future gas reserves are predicted to come from the Brent Group. Pressure-depletion rate is one of the important factors influencing the gas reserves. Adequate modeling of vertical lift performance (VP) is needed to predict a realistic liquid offtake and thereby pressure-depletion rate. Main objective of this thesis is to give a recommendation of which correlation(s) to be used when generating VP curves for the Statfjord full field model. Furthermore, modifications of correlations and how this may affect simulation result are studied to give a recommendation of how correlations should be used when generating lift curves. The correlation(s) recommended should give good result over the range of 1

10 production conditions expected regarding drainage strategy, and modifications should not introduce errors that may increase with time. Production tests from selective wells located in the Statfjord Formation have been used as basis when investigating the different correlations. Calculations are performed using Prosper, a well performance, design and optimization program developed by Petroleum Experts (2010). Wells in the Statfjord Formation have been producing from a gas cap which in some areas of the formation has disappeared. These wells have therefore been producing with various gas-liquid ratios (GR), one of the main parameters influencing lift. Wells in the Brent Group have currently low GR, but with depressurization GR will increase. The correlation s accuracy when predicting bottomhole pressures with varying GR is studied. Furthermore, modifications of correlations and the effect on simulation results are studied. 2

11 2 Theory 2.1 Single-Phase Flow Single-phase flow is unexpected in a producing well. Even if only one phase is produced from the reservoir, pressure depletion across the pipeline may generate multiphase flow (Time 2009). Before heading into multiphase flow, a general understanding of single-phase flow is useful. Calculations for single-phase flow act as basis for multiphase flow (Brill and Mukherjee 1999). In this section, the steadystate pressure-gradient equation for single phase flow will be described. The different terms will be discussed, and a brief description of laminar and turbulent flow is given Pressure-Gradient Equation The steady-state pressure-gradient equation is found by combining equations for conservation of mass, dp dt ( v) 0,....(2.1) and linear momentum, ( v) ( v t 2 p d ) A g cos,. (2.2) where p is pressure, t is time, ρ density, v velocity, length, τ shear stress, d pipe diameter, A pipe cross-sectional area, g gravity and θ inclination angle from vertical. By assuming steady-state flow the pressure-gradient equation may be expressed as, dp d d dv g cos v. (2.3) A d As seen from Eq. 2.3, the total pressure gradient in a pipeline may be expressed as the sum of three components, 3

12 dp d t dp d f dp d h dp d,...(2.4) a where (dp/d) t is total pressure gradient, (dp/d) f frictional pressure gradient, (dp/d) h hydrostatic pressure gradient and (dp/d) a acceleration pressure gradient (Brill and Mukherjee 1999). When calculating frictional pressure drop in single-phase flow, it is important to discriminate between laminar and turbulent flow. The type of flow is determined from Reynolds number N Re vd, (2.5) where N Re is Reynolds number and µ is viscosity. One may discriminate between flow regimes the following way: N Re 2000: aminar flow 2000 < N Re 4000: Transition between laminar and turbulent flow N Re > 4000: Turbulent flow In single-phase laminar flow, at constant flow velocity and pipe diameter, the frictional pressure drop is given by: dp d f 4 d v, (2.6) N Re 2 if the Fanning friction factor is used. The Moody type friction factor may also be used for laminar flow. Then the number 4 in equation 2.6 is included in the friction factor term, 4

13 dp d f 1 d v....(2.7) N Re 2 The result will be exactly the same whether Fanning friction factor 16 f F,... (2.8) N Re or Moody friction factor 64 f M,... (2.9) N Re is used. For laminar flow, the friction factor may be determined exactly from the theory, due to the well defined parabolic velocity profile. In turbulent flow, the velocity profile becomes more uniform, although fluctuating. arger velocity fall-off towards the pipe wall results in a larger shear rate. Thereby various equations exist for calculating turbulent friction factors (Time 2009). For smooth pipes, and high Reynolds number one may use the following equation to determine the friction factor: f CN n Re, (2.10) where C and n are correction factors. The correction factors are found experimentally. When C = and n = -0.25, Time (2009) refers to equation 2.10 as the Blasius form. The pipe wall is normally not smooth, and one must account for the wall roughness. In turbulent flow the friction factor has been found to depend on relative roughness and the Reynolds number. Brill and Mukherjee (1999) write that Nikuradse rough pipe friction factor correlation, 5

14 1 f log, (2.11) d is based on the relative roughness (ε/d). If the friction factor varies both with Reynolds number and relative roughness, the region is defined as transition or partially rough wall. Colebrook friction factor, log f d N 18.7 Re f, (2.12) is made to describe the variation of friction factor in the transition region (Brill and Mukherjee 1999). For fully developed turbulent flow (rough pipe flow) with large Reynolds number, Eq degenerates to Eq A Moody diagram is often used to find friction factors. For a given Reynolds number and relative roughness one may read a friction factor. The Moody diagram in figure 2.1, shows the variations of friction factors based on Eq and the friction factor in laminar flow (Brill and Mukherjee 1999; Time 2009). Figure 2.1: Moody Diagram (Brill and Mukherjee 1999) 6

15 The gravity term is given by the fluid density and the pipe inclination relative to the vertical direction (θ): dp d h g cos....(2.13) Contribution from the acceleration term is normally insignificant. Acceleration may give a relatively small contribution if the velocity of the producing fluid changes rapidly (e.g. in a gas well operating at low wellhead pressure) (Time 2009). Rapid change in velocity may cause a pressure change, and contribute to the total pressure gradient. The acceleration pressure gradient (one dimensional) is given as (Brill and Mukherjee 1999): dp d a v dv d....(2.14) 7

16 2.2 Multiphase Flow Multiphase-flow behavior is much more complex than single-phase flow. If gas and liquid flows simultaneously, they tend to separate because of differences in fluid properties. The fluids will give different shear stress due to differences in density and viscosity. Gas and liquid will normally not travel with the same velocity. In vertical flow the gas phase tends to have a higher velocity compared to the liquid phase. This occurs because gas is more compressible, less dense and less viscous than liquid (Brill and Mukherjee 1999). There will be several forces acting on the fluids, buoyancy, turbulence, inertia and surface tension. The relative magnitude of these forces may change along the pipe, resulting in different flow regimes (Brennen 2005). To deal with the complex nature of multiphase flow, many flow parameters and various mixing rules are defined. These make it possible to use the same basic pressure gradient equation as for single phase flow, modified for multiphase flow. The basic definitions of flow parameters, flow patterns and general equations for mixing are presented in this section Holdup The proportion of the pipe cross-section or volume that is occupied by the liquid phase is defined as the liquid holdup (H ) (Brill and Mukherjee 1999). Experimentally it is found by averaging liquid or gas volume versus total amount of fluid, see equations , (a = gas or liquid). If gas is used, the liquid holdup is found as, H 1 H ),...(2.15) ( G because the sum of fraction gas and liquid should be one. Generally it is discriminated between line-average H a a,...(2.16) area-average 8

17 H a A a A,...(2.17) and volume-average V H a a V,...(2.18) given the averaging-measurement method used. If the mixture was completely homogeneous, the three methods should give the same fluid fractions (Time 2009). It may be difficult to measure the fluid fractions, e.g., in a subsea pipeline. Estimation of the liquid holdup then becomes crucial. If the volumetric flow rates (q, q G ) are known, the no-slip fractions (flux fraction) may be calculated. No-slip liquid fraction (λ ) is given by, q,...(2.19) q qg and no-slip gas fraction (λ G ) as, q G G....(2.20) qg q If the phase velocities are different, slip is present. Gas has higher mobility compared to liquid, giving gas a higher velocity. The ratio between the real phase velocities defines the slip ratio, v G S....(2.21) v Slippage of gas past liquid results in larger liquid holdup, compared to the situation of no-slip. If slip is present, the fluid fractions cannot be calculated using Eq and If the slip ratio is known, the true fluid fractions may be determined using 9

18 H for liquid and A A q,...(2.22) A A A 1 G q qg S H G A G G G,...(2.23) A A G A A q G q Sq for gas. From Eq and 2.23, H = λ and H G = λ G if S = 1 (Time 2009). How liquid holdup is estimated in multiphase-flow correlations varies amongst the authors. This will be described in greater detail in section Velocities Superficial velocity is defined as the velocity of a phase if it was occupying the entire pipe area. Superficial velocity (v S ) for liquid is given by: q v S A,...(2.24) and for gas qg vsg....(2.25) A The real average velocity in a pipe, defined as the total mixture velocity (v m ), may be found as the sum of superficial velocities, v m q q G vs vsg....(2.26) A Real phase velocities may be defined locally or as time- and space-averaged velocities. If holdup is known, one may determine the real flowing cross sections for liquid and gas, and thus the real phase velocities by: 10

19 q v,...(2.27) A and q G vg....(2.28) AG The difference between the real gas velocity and the real liquid velocity is defined as the slip velocity (v s ) (Time 2009), v s v v....(2.29) G Mixture-Fluid Properties Multiphase-flow correlations in general consider only two phases, liquid and gas. Water and oil may be combined and treated as one equivalent fluid and referred to as the liquid phase (Petroleum Experts 2010). In this thesis mixing rules for oil and water will not be included, but there exist various ways to combine water and oil to one fluid. Many equations to calculate fluid properties for a mixture of gas and liquid have been proposed. If the equations consider slip or no-slip fractions is the main difference between them. Mixture density with slip (ρ m ) can be found by, H 1 H )....(2.30) m G ( Mixture density with no-slip (ρ mn ) is found by replacing H with λ (Brill and Mukherjee 1999): mn 1 )....(2.31) G ( 11

20 Several models to determine mixture viscosity exist. These arise because mixture viscosity is strongly dependant on dynamical processes, including bubble size, flow regime etc (Time 2009). The most common equations are listed below: H 1 H )...(2.32) m G ( H ( 1H )...(2.33) m G 1 )...(2.34) mn G ( Pressure-Gradient Equation Pressure-drop calculation for two-phase flow is quite similar that of single-phase flow. The main difference is the use of mixed fluid properties for two-phase flow. The total pressure-gradient equation takes the same form as for single-phase flow, Eq Each term is modified for two phases, and is described in the following section. Frictional pressure drop may be expressed as (Brill and Mukherjee 1999): dp d f 4 D f 1 mv 2 2 m....(2.35) Various two-phase friction factors, and properties used when calculating the Reynolds number varies amongst authors. This will be described for different correlations in section 2.4. Pressure drop caused by the hydrostatic term is normally the larges contribution to the total pressure drop, for wells producing liquid. For conditions of high gas velocities, the frictional pressure drop may exceed the contribution from the hydrostatic term. Pressure drop caused by the hydrostatic term is found by: dp d h g cos....(2.36) m 12

21 The mixture density is usually calculated using Eq (Brill and Mukherjee 1999). How the liquid holdup is correlated thus becomes crucial for the hydrostatic pressure drop. As mentioned for single-phase flow, the acceleration pressure drop is normally negligible. It is considered mainly for cases of high fluctuating flow velocities. For two-phase flow the pressure-drop component caused by acceleration can be found from (Brill and Mukherjee 1999): dp d a v m m dvm d....(2.37) Flow Regimes Single-phase flow is divided into laminar and turbulent flow regimes. In multiphase flow the discrimination becomes more complex. Gas and liquid distribution may vary when flowing in a long pipe, resulting in different flow regimes (Time 2009). A brief description of the flow regimes that may occur in vertical flow will be given in this section. In general one may discriminate between four flow regimes for vertical upward multiphase flow: bubble flow, slug flow, churn flow and annular flow, see figure 2.2. The flow regimes change in this order by increasing gas rate for a given liquid rate (Zavareh et al. 1988). The most important flow patterns for multiphase flow in wells are slug and churn flow patterns. They are often referred to as intermittent flow regimes (Brill 1987). Mist flow and annular-mist flow are other names for the annular flow regime (Brill and Mukherjee 1999). 13

22 Figure 2.2: Flow patterns for upward vertical flow (Brill 1987) In bubble flow, liquid is the continuous phase and the free-gas phase is presented as small bubbles. The gas-bubbles are randomly distributed in the liquid flow, and the diameter may vary. Due to different sizes of the gas-bubbles, they travel with different velocities. The liquid phase however moves with a more uniform velocity. The gas phase, except for its density, has little effect on the pressure drop (Orkiszewski 1967). Slug flow is characterized by alternating slugs of liquid with large bubbles of gas. arge gas-bubbles are made as the smaller gas-bubbles coalesce, when gas velocity increases. The larger bubbles are called Taylor bubbles. Smaller bubbles of gas are contained in the liquid slugs. iquid is still the continuous phase, because of a liquid film covering the Taylor bubbles (Zavareh et al. 1988). As the gas velocity is increased further, the large gas-bubbles become unstable and may collapse. When this happens, churn flow occur. Churn flow is a highly turbulent and chaotic regime. Neither gas nor liquid phase appears to be continuous. Oscillatory, up and down motion of liquid, is characteristic for churn flow (Zavareh et al. 1988). 14

23 In annular flow, gas is the continuous phase. Gas flows with a high rate in the centre of the pipe. iquid is found as a liquid film coating the pipe wall and as entrained droplets in the gas phase. The gas phase becomes the controlling phase (Orkiszewski 1967). Determination of flow regime will be important for parameters such as holdup and thereby pressure-drop predictions. Result of study on flow regimes are often displayed in the form of a flow regime map (Brennen 2005). Flow maps are generated to relate flow patterns to flow rates and fluid properties. Boundaries in a flow regime map represents where a regime becomes unstable. A growth of the instability will lead to transition to another regime. These transitions can be rather unpredictable because they may depend on otherwise minor features of the flow, as the wall roughness or entrance conditions. Hence, the flow-pattern boundaries are not distinctive lines, but more poorly defined transition zones. Many different flow regime maps have been published, based on different correlations for flow-regime prediction. Most of them are dimensionless and applies only for the specific pipe size and fluids used when they were created (Brennen 2005; Zavareh et al. 1988). 15

24 2.3 Calculation of Pressure-drop in ong Pipelines The pressure will drop when fluids flow from inlet to outlet in a long pipeline. The gas density, and thereby the gas velocity will change according to the pressure changes. As the pressure drop, more and more gas may evaporate from the oil phase into the gas phase. This will in turn increase the gas flow velocity, which again will lead to higher pressure drop and even higher evaporation. By this, the pressure at inlet and outlet determines the total flow velocity. At least one more factors complicate the calculations, namely temperature. A temperature profile along the pipeline and the heat conduction from the surroundings are needed to determine the pressure traverse (Time 2009). The total pressure drop over a pipeline may be calculated by integrating the pressure gradient over the total length, p 0 dp d d....(2.38) The challenge lies in the fact that the pressure gradient is dependent on pressure, temperature and inclination angle, and will vary throughout the pipe length. Properties like flow pattern, densities, rates etc. will be affected. A general approach is to divide the pipeline into segments, and calculate pressure stepwise along the pipe. The segments should be small enough so that the pressure gradient can be considered constant within the segment. If the flow rates of oil and gas and the inlet pressure are known, it is possible to calculate the pressure at the outlet. Calculations may also be carried out the other way around. Pressure is calculated stepwise, and the flow rates are updated for each segment along the pipe. The pressure gradient is calculated for each segment and multiplied by the length of the segment, p n i1 dp i d....(2.39) i 16

25 The outlet pressure for segment i will be the inlet pressure for segment i+1. Pressure obtained at the end of the last segment will be the outlet pressure. The total pressure drop will be the sum of pressure drops calculated for each segment (Time 2009; Brill and Mukherjee 1999). 17

26 2.4 Pressure-Drop Correlations A large range of different pressure-drop correlations are published. In addition many methods and correlations developed are kept confidential. As stated in by Time (2009); There is no guarantee that the correlations kept confidential are better than other correlations. On the contrary, keeping methods secret is a way to avoid scientific testing, and the methods may have low validity. One may divide the pressure gradient calculations into two categories: 1) Empirical correlations, based on experimental data and dimensional analysis. 2) Mechanistic models, based on simplified mechanistic (physical) considerations like conservation of mass and energy. It can be quite difficult to discriminate between empirical and mechanistic correlations. Often a combination is used to develop multiphase correlations (Yahaya and Gahtani 2010). The empirical correlations are generated by establishing mathematical relations based on experimental data. Dimensional analysis is often used to select correlating variables. It is important to notice that application of empirical correlations is limited to the range of data used when it was developed (Ellul et al. 2004; Yahaya and Gahtani 2010). Further it is possible to divide the empirical correlations in groups regarding if slip and flow patterns are considered, see table 2.1. The mechanistic models are based on a phenomenological approach and they take into account basic principles, like conservation of mass and energy (Yahaya and Gahtani 2010). In mechanistic models, flow regime determination is important. Normally a mechanistic transport equation is written for each of the phases in the multiphase flow. Separate models for predicting pressure drop, liquid holdup and temperature profile have been developed by flow regime determination and separating the phases (Ellul et al. 2004). 18

27 Table 2.1: Classification of correlations Correlation Category Slip considered? Flow regime considered? Fancher & Brown Empirical No No Gray Empirical Yes No Hagedorn & Brown Empirical Yes No Duns & Ros Empirical Yes Yes Orkiszewski Empirical Yes Yes Beggs & Brill Empirical Yes Yes Mukherjee & Brill Empirical Yes Yes Petroleum Experts (1,2,3) Empirical Yes Yes Petroleum Experts (4,5) Mechanistic Yes Yes Hydro 3-Phase Mechanistic Yes Yes OGAS Mechanistic Yes Yes Most correlations defined as empirical in table 2.1 will be described regarding theory. In the experimental work, a few mechanistic models are used. These are the once listed in table 2.1, and will not be described here. Similar equations for pressure drop are proposed for the empirical correlations. The main difference between the correlations is how liquid holdup, mixture density and friction factors are estimated. Descriptions of the various estimations for the respective correlations are found in the following sections Fancher and Brown Correlation (Fancher and Brown 1963) Fancher and Brown proposed a correlation based on Poettmann and Carpenter s (1952) work. As table 2.1 describes, this is a no-slip correlation and the same correlation is used regardless of flow regime. An 8000 ft long experimental field well with 2 3/8 inch OD tubing was used for testing. Flow rates ranged from B/D at various GR from 105 to 9433 scf/bbl. Results from these tests were compared with Poettmann and Carpenter s (1952) correlation. Deviations occurred for low flow rates and for high GR (outside the range of what Poettmann and Carpenter s correlation was designed for). The deviation was believed to originate from the 19

28 friction factor correlation. By adopting the pressure gradient equation from Poettmann and Carpenter (1952), a new correlation for friction factor was proposed. The pressure gradient equation may be expressed as, dp dz 2 f mn vm 2d mn g,.(2.40) for vertical flow of a homogeneous no-slip mixture (Brill and Mukherjee 1999). Fancher and Brown found that GR is a significant parameter in the friction factor correlation. Thereby three separate friction factor correlations were developed, divided by GR ranges of Mscf7bbl, Mscf/bbl and greater than 3.0 Mscf/bbl, see figures 2.3 to 2.5. Figure 2.3: Friction factor correlation (Fancher and Brown 1963) Figure 2.4: Friction factor correlation (Fancher and Brown 1963) 20

29 Figure 2.5: Friction factor correlation (Fancher and Brown 1963) Gray Correlation (Gray 1974) Gray developed an empirical correlation for a vertical well producing gas and gascondensate or water. Slip is considered, but it does not distinguish between different flow patterns, see table 2.1. The correlation is based on a total of 108 well test data, and Gray cautioned use of the correlation beyond the following limits: velocities higher than 50 ft/sec nominal diameters larger than 3.5 in condensate or liquid loadings above 50 bbl/mmscf water or liquid loadings above 5 bbl/mmscf Pressure gradient for two-phase flow is given by: dp dz f 2 mnvm d m g mnv m 2d dz mn....(2.41) With basis in dimensional analysis, Gray s correlation uses three dimensionless parameters to predict liquid holdup, namely velocity number N v 2 4 mnvm,...(2.42) g ( ) G where σ is surface tension. Dimensionless diameter number 21

30 N D g( ) d 2 G,...(2.43) and superficial liquid-gas ratio v S R....(2.44) v SG Gray proposed the following equation for predicting liquid holdup, B Exp 2.314N 1 v 1 N D H,...(2.45) R 1 where 730R B ln1....(2.46) R 1 The three dimensionless parameters defined above are intersected into equation 2.45 to estimate liquid holdup. When liquid holdup is determined, mixture density may be calculated using equation If both condensate and water are present, Gray suggested that surface tension should be calculated as f f o o w w,...(2.47) f o f w where f o,w is no-slip volume fraction for condensate and water. The friction loss model is a modified Darcy-Weisbach expression, and the flow is assumed to be turbulent. By this the energy loss is considered wholly dependent on a 22

31 pseudo wall roughness factor (ε). A Colebrook-White function together with pseudo wall roughness factor is used to obtain a two-phase friction factor. The pseudo wall roughness factor is correlated using a roughness variable defined by a modified Weber number. If R 0.007, pseudo wall roughness is given by ' 28.5,...(2.48) 2 mnvm where ε is a roughness variable. If R< 0.007, one should use G ' G R 0.007,....(2.49) when calculating pseudo wall roughness. The two-phase friction factor may then be read of a Moody diagram, see figure 2.1. By definition ε must be larger or equal to 2.77x10-5. In Prosper a modified version of the Gray correlation is used. It was modified by Shell, but no paper documenting the modifications was found Hagedorn and Brown Correlation (Hagedorn and Brown 1965) The Hagedorn and Brown correlation is in the same category as the Gray correlation, see table 2.1. To develop the correlation an experimental vertical well of 1500 ft was used. The pressure gradient occurring during continuous two phase flow was studied in tubing with 1 in., 1 ¼ in. and 1 ½ in. nominal diameter. Air was used as the gasphase. The liquid phase was varied. Water and crude oils with viscosities approximately 10, 30 and 110 cp were used. iquid flow rates and GR were also varied between the tests. During development of the correlation, Hagedorn and Brown did not measure the liquid holdup. They developed a pressure-gradient equation, and by assuming a friction factor correlation they could calculate pseudo liquid holdup values to match 23

32 measured pressure gradients. The correlation for liquid holdup is therefore not based on true measurements of liquid holdup. The pressure-gradient equation developed has the form: dp dz 2 f mnv 2 d m 2 m m v m g 2dZ 2 m....(2.50) The holdup correlation is shown in figure 2.6. In order to determine holdup, a secondary correction factor (ψ), and a corrected liquid-viscosity number (CN ) must be determined. These factors are found from figures 2.7 and 2.8 by using dimensionless groups proposed by Duns and Ros (1963). The dimensionless groups are liquid velocity number N v g 4 V S,...(2.51) gas velocity number N v g 4 GV SG,...(2.52) pipe diameter number g N D d,...(2.53) and liquid viscosity number g N 4....(2.54) 3 When holdup is determined, the mixture density may be calculated using Eq

33 Figure 2.6: Holdup-factor correlation (Hagedorn and Brown 1965) Figure 2.7: Correlation for liquid viscosity number (Hagedorn and Brown 1965) Figure 2.8: Correlation for secondary correction factor (Hagedorn and Brown 1965) Darcy-Weisbach equation for single phase flow, relative roughness of the pipe and the two-phase Reynolds number are used to determine the two-phase friction factor from a Moody diagram, see figure 2.1. When calculating the Reynolds number, an assumption stating that the mixture of gas and liquid can be treated as a homogenous mixture over a finite interval is used. The Reynolds number for the two phase mixture may than be written as 25

34 N v d m mn Re,...(2.55) TP m where µ m is defined by equation Modification has been proposed to the Hagedorn and Brown correlation. The refinements suggested by Brill and Hagedorn have been implemented in Prosper (Petroleum Experts 2010): Griffith correlation for bubble flow imit on liquid holdup to always be greater than the no-slip holdup Some additional refinements have been added to the basic Hagedorn and Brown correlation in Prosper (Petroleum Experts 2010): Beggs and Brill deviation correction for liquid holdup Explicit calculation of acceleration term Duns and Ros Correlation (Duns and Ros 1963) The Duns and Ros method is an empirical correlation based on approximately 4000 two-phase flow experiments. iquid holdup and pressure gradients were measured. The experiments were conducted as vertical flow, with pipe diameters ranging from 1.26 to 5.60 inches. Flow patterns were observed in a transparent section of the test tubing. In the Duns and Ros correlation it is discriminated between three main flow regimes. iquid holdup and friction factor correlations were developed for each flow regime. Duns and Ros correlation discriminates between three different flow regimes. These are shown in figure 2.9, described as regions. In region I, liquid is the continuous phase. Where gas and liquid phase s alternate is referred to as region II and in region III gas is the continuous phase. A transition regime is treated as a fourth regime in calculations. For flow in the transition regions linear interpolation may be used to approximate the pressure gradient. 26

35 N V N GV Figure 2.9: Flow regime map (Duns and Ros 1963) Both friction factor and liquid holdup were found to depend on gas and liquid velocities, the pipe diameter and the liquid viscosity. These factors together with surface tension and liquid density are converted into four dimensionless groups as described in Eqs to Duns and Ros used a dimensionless slip velocity number, S l v 4 s,...(2.56) g to correlate liquid holdup, H v v v s s m m s s S....(2.57) 2v v 2 4v v 27

36 Determination of slip varies between the three regions, one correlation for each region. The correlations are based on liquid and gas velocity numbers and a given number of constants related to viscosity and diameter. For a more detailed description and equations it is referred to the original paper. When slip is determined from a respective correlation, Eq is solved for slip velocity. Furthermore liquid holdup is calculated using Eq and mixture density may be calculated using Eq The hydrostatic pressure gradient may be calculated as described in Eq In region I and II, the pressure-gradient due to friction is found from dp dz f vsv d f 2 m....(2.58) The friction factor correlation developed by Duns and Ros is based on experimental data. The following equation was proposed, f f f,...(2.59) 2 1 f 3 where f 1 is a function of the Reynolds number for liquid, and may be found from figure Besides the transition region between laminar and turbulent flow, figure 2.10 is identical to the Moody diagram for single phase flow. The factors f 2 and f 3 are correction factors for in-situ GR, and both liquid viscosity and in-situ GR respectively. Figure 2.10: Non-dimensional f 1 versus Reynolds number (Duns and Ros 1963) 28

37 In region III friction is assumed to originate from the drag of gas on the pipe wall. Due to this assumption the friction gradient is based on the gas phase, dp dz fgv d f 2 2 SG....(2.60) No slip gives f = f 1, and the Reynolds number should be calculated for the gas flow. The friction factor may then be read of figure Duns and Ros found that wall roughness actually is the roughness of the liquid film. Ripples in the liquid film are formed due to drag of gas, thus roughness will not be constant. They suggested a way to account for this effect. It is referred to original paper for details. In Prosper the following refinements have been made to the basic Duns and Ros method (Petroleum Experts 2010): Beggs and Brill deviation correction for holdup Gould et al. (1974) flow map Explicit calculation of the acceleration term Orkiszewski Correlation (Orkiszewski 1967) Orkiszewski compared many of the published correlations against test data. He concluded that none of them sufficiently described two phase flow for all the flow regimes. Thereby a combination of the correlations that best described the test data was suggested to be used. Orkiszewski uses Griffith and Wallis method for slug flow, Duns and Ros for transition and mist flow, and he suggested a new method for slug flow. Determination of flow regime is described in table 2.2. Griffith and Wallis have defined the boundary between bubble and slug, while Duns and Ros have defined the boundaries for the remaining three regimes. The variables are described in equations 2.61 to

38 Table 2.2: Flow-regime boundaries for Orkiszewski correlation Flow Regime imit Bubble v SG /v m < B Slug v SG /v m > B, N GV < S Transition M > N GV > S Mist N GV > M Bubble-slug boundary ( B ) is defined by B 2 vm ,...(2.61) d with the constrain B Slug-transition boundary, ( S ) are given as 36N q S 50 VG,...(2.62) qg And transition-mist boundary ( M ) is 0.75 vgq M....(2.63) qg In bubble flow liquid holdup given is given by: H v 1 v m s v 1 v m s 2 4v v SG s....(2.64) According to Orkiszewski, Griffith suggested an average value of the slip velocity to be used as a constant equal to 0.8 ft/sec. The average flow density is found from Eq. 30

39 2.30, together with the liquid holdup the hydrostatic pressure-gradient may be calculated as described in equation The friction pressure-gradient is given by dp dz f vs H d f (2.65) Friction factors are obtained from a Moody diagram using liquid Reynolds number, N v / H d S Re,...(2.66) and relative roughness. The slip density for slug flow proposed by Orkiszewski is: m v v v S b G SG,...(2.67) vm vb where v b is bubble rise velocity and Γ is a liquid distribution coefficient. Γ is correlated from oilfield data by Hagedorn and Brown (1965) as described in table 2.3. Bubble rise velocity is defined as v b 1 2 C C gd....(2.68) Here C 1 and C 2 are expressed as a function of bubble Reynolds number N v d b Re B,...(2.69) and liquid Reynolds number, 31

40 N v d m Re....(2.70) After calculating liquid and bubble Reynolds numbers, C 1 and C 2 may be read off figures 2.11 and 2.12 respectively. Figure 2.11: Griffith and Wallis C 1 versus Reynolds number (Orkiszewski 1967) Figure 2.12: Griffith and Wallis C 2 versus bubble Reynolds number and Reynolds number (Orkiszewski 1967) The friction pressure-gradient may be found from dp dz f f v 2d 2 m v v S m v v b b....(2.71) 32

41 Friction factor is obtained using a Moody diagram and liquid Reynolds number. The liquid distribution coefficient may be found as described in table 2.3 together with respective equations. Table 2.3: iquid distribution coefficient equations Continuous liquid phase v m Use equation number Water < Water > Oil < Oil > log / d logv 0.428log d.. (2.72) m log / d log v 0.888log d. (2.73) m log( 1) / d log v 0.113log d..(2.74) log( 1) / d log v m 0.01log( 1) / d m log d...(2.75) log d The liquid distribution coefficient is constrained by the limit 0.065v m, (2.76) if v m < 10 ft/sec, and v m v b v b m 1,.(2.77) 33

42 when v m > 10 ft/sec. The constraints are made to eliminate pressure-discontinuities between the flow regimes. Still significant discontinuities may occur (Petroleum Experts 2010). For transition and mist flow, the correlations developed by Duns and Ros are to be used, see section Beggs and Brill Correlation (Beggs and Brill 1973) Beggs and Brill developed correlations for liquid holdup and friction factor. The correlations are based on experimental data from 90 ft long acrylic pipes. Fluids used were air and water and 584 tests were conducted. Gas rate, liquid rate and average system pressure was varied. Pipes of 1 and 1.5 inch diameter were used. First the pipe was horizontal, and the flow rates were varied in such a way that all horizontal flow patterns were observed, see figure Afterwards the pipe inclination was changed, and liquid holdup (H (θ) ) and pressure drop was measured. By this the effect of inclination on holdup and pressure drop could be studied. Beggs and Brill proposed the following pressure-gradient equation, dp d 2 f nvm m g sin 2d, (2.78) 1 E k where E k, dimensionless kinetic-energy pressure gradient, is defined by E k v m v SG p n and mixture density should be calculated as,...(2.79) m H H ( ) G 1 ( )....(2.80) iquid holdup and friction factor should be found as described in the following. 34

43 Figure 2.13: Horizontal flow patterns (Beggs and Brill 1973) Beggs and Brill plotted liquid holdup versus angle of pipe from horizontal, see figure They found that holdup has a definite dependency on angle. From the figure one can see that the curves have maximum and minimum at +/- 50 from the horizontal. The slippage and liquid holdup increase as the angle of the pipe increase, from horizontal towards vertical (flow upwards). Gravity forces act on the liquid, causing a decrease in the liquid velocity and thereby slippage and holdup is increased. By further increasing of the angle, liquid covers the entire cross section of the pipe. The slippage between the phases is reduced and liquid holdup reduces. Beggs and Brill observed that degree of holdup with angle varied with flow rates. To include effects of pipe inclination, it was decided to normalize liquid holdup. The following equation was proposed, H H 0,..(2.81) where Ψ is inclination correction factor, H (θ) is holdup at angle θ from horizontal, and H (0) is horizontal holdup. 35

44 Figure 2.14: iquid holdup versus angle (Beggs and Brill 1973) The liquid holdup for horizontal flow should be calculated first, and corrected for inclination afterwards. The equations used for calculating liquid holdup is the same for all flow patterns, but there are different empirical coefficients for each flow pattern. The equation for calculating liquid holdup for horizontal flow is: H a b ( 0),..(2.82) N c Fr where a, b, c are empirical coefficients given in table 2.4 and N Fr is mixture Froude number N Fr 2 vm. (2.83) gd Table 2.4: Empirical coefficients for calculating liquid holdup Flow Pattern a b c Segregated Intermittent Distributed iquid holdup for horizontal flow should be grater or equal to the no-slip liquid volume fraction. The inclination correction factor is given by, C sin(1.8 ) 0.333sin (1.8 ),.(2.84) 36

45 where θ is actual angle of pipe from the horizontal, and C is liquid holdup parameter. The liquid holdup parameter is defined as f C (1.0 )ln( e N g V N h Fr ),.(2.85) with a restriction, C 0. e, f, g, h are empirical coefficients. They vary with flow regime and flow direction and should be determined from table 2.5. Only uphill flow direction is included here. For the distributed flow pattern no correction is needed. C will be zero, giving ψ equal to one. If the flow falls in the transition regime, an interpolation should be carried out. Table 2.5: Empirical coefficients for calculating liquid holdup parameter Flow Pattern e f g h (Uphill) Segregated Intermittent Values for the two phase friction factor were found by solving the pressure-gradient equation, Eq The two-phase friction factor was normalized by dividing it by a no-slip friction factor (f n ). This may be found from a Moody diagram, see figure 2.1 or by using no-slip values in Reynolds-number and equations for smooth pipe friction factor. Based on these values, Beggs and Brill proposed the following equation for two phase friction factor, f f n s e,.(2.86) where S ln y ln y (ln y) (ln y) 4, (2.87) and 37

46 y H 2 ( ). (2.88) To ensure that the correlation degenerates to single phase flow when y = 1, Beggs and Brill proposed that S should be calculated as, S ln( 2.2y 1.2),. (2.89) when 1 < y < Petroleum Experts Correlations (Petroleum Experts 2010) Petroleum Experts correlations are a combination off different correlations. It is developed by the company Petroleum Experts. Papers describing the correlations have not been found, only a brief description in the Prosper manual. It is believed that the correlations are used as described earlier. Flow regimes are determined using Gould et al. (1974) flow map. See table 2.6 for correlations used in the various flow regimes. iquid holdup and frictional factors are found using the respective flow correlations. Table 2.6: Correlations used by Petroleum Experts correlations Flow regime Correlation Bubble Flow Wallis and Griffith Slug Flow Hagedorn and Brown Transition Duns and Ros Annular Mist Flow Duns and Ros It is stated that by using the physical properties of vapor and liquid phases determined in situ, the individual flow rates, pipe geometry and pressure level, one can predict the flow regime likely to occur at that particular point. Gould et al. (1974) have developed a flow regime map for vertical flow, see figure It was developed on the basis of results from observations on literature data and laboratory experiments. 38

47 Figure 2.15: Flow regime map (Gould et al. 1974) The regimes are discriminated into liquid-phase continuous (bubble flow), alternating phases (slug flow), gas-phase continuous (mist flow), transition (between slug and mist) and a heading region (both phases continuous). It is important to recognise that the location of the flow regime boundaries only is approximated (Gould et al. 1974). 39